Support for Drupal 7 is ending on 5 January 2025—it’s time to migrate to Drupal 10! Learn about the many benefits of Drupal 10 and find migration tools in our resource center.
I am in the process of writing support for installs with remote DB servers. I will then move on to upgrade support.
This involves hacking a bit more install.hostmaster.inc, migrate.hostmaster.inc and the profile.
This could help resolve #952144: support local socket connexions for mysql.
I think this should part of the release so i am marking this critical.
Comments
Comment #1
anarcat CreditAttribution: anarcat commenteddealing with installs first.
I have pushed the branch dev-multiserver-install for provision which does the first half (the backend).
for the second half, I wonder if I should check master_db or if we shouldn't just drop master_db and rely on the dbserver associated with the @hostmaster site?
either way, the frontend profile needs to be changed to create a new node if the dbserver is different from the master server.
this should be around hostmaster.profile:77.
Comment #2
anarcat CreditAttribution: anarcat commentedi guess what I mean regarding master_db is that it doesn't seem to be the right way to pass that information around. In fact, I believe it *shouldn't* be passed to the frontend at all in this case. But maybe that would break the edition of this server.
Either way, I've always been quite unconfortable with storing the mysql root password(s) in the frontend DB. But that's a separate issue.
Comment #3
anarcat CreditAttribution: anarcat commentedFWIW, i have setup a new hostmaster instance with the above patches and the only problem is that the DB server is not created in the frontend. The install is otherwise fully functional and manually creating the node fixes the problem completely.
Comment #4
anarcat CreditAttribution: anarcat commentedCreated the branch on hostmaster, we now create the DB node if it's different from master. I just need some testing now.
Comment #5
anarcat CreditAttribution: anarcat commentedSo I have now completed the frontend: the last commit had a small bug... But now a node is created for the DB server in the frontend if it doesn't already exist.
I have tested this repeatedly with the following:
I would like somebody to test this and the "normal" install. miguel? Once that is proven as working in most cases, mark this as RTBC and i'll merge on master.
Thanks!
Comment #6
sfyn CreditAttribution: sfyn commentedI am attempting an install in a vbox with some unusual symptoms, using hostmaster-install:
drush hostmaster-install --aegir_db_host=localhost --working-copy --debug
And I am consistently generating the same errors:
Vhost configuration and such appear to be completing correctly, except that for whatever reason, I am consistently getting garbage characters inserted at the beginning of my server url.
Any thoughts?
Comment #7
sfyn CreditAttribution: sfyn commentedStrangely enough this error appears to have gone away on its own...
Apparently it is sufficient to do an rm-r of the existing sites file after having run drush provision-delete to be able to safely reinstall.
Comment #8
anarcat CreditAttribution: anarcat commentedThe PROVISION_SITE_INSTALLED error means what it says: the hostmaster site is already there. :) Run drush @hostmaster provision-delete first.
Comment #9
anarcat CreditAttribution: anarcat commentedSo this was tested with --aegir_db_host, we need someone to test the regular install.
Comment #10
Anonymous (not verified) CreditAttribution: Anonymous commentedI shall, but won't get the chance for 24 hours sorry.
Maybe Grace if you are in a hurry
Comment #11
omega8cc CreditAttribution: omega8cc commentedI will try that today.
Comment #12
anarcat CreditAttribution: anarcat commentedThank you grace!!!
Comment #13
sfyn CreditAttribution: sfyn commentedJust for the record, I did use drush @hostmaster provision-delete first.
Comment #14
anarcat CreditAttribution: anarcat commentedI am able to install without the --aegir_db_host argument too, on my usual test platforms. ie. I do not think this is adding new regressions.
I think I will be merging this in after one last pass at fixing prompts and workflow.
Comment #15
anarcat CreditAttribution: anarcat commentedI have merged the branch in master now. I think this is now complete: upgrades should just follow because nodes and contexts are created properly.